The phenomenon of alternate bearing of fruits seriously affects the fruit yields as well as the economic benefits of orchards. The present study investigated the possibility of airborne hyperspectral images to predict the fruit yield of individual citrus trees. The hyperspectral data were first extracted from the images and the predictors were determined using partial least-squares regression (PLS). The optimal number of PLS factors were identified, and they were used as inputs of citrus yield prediction models developed by means of multiple linear regression (MLR) and artificial neural network (ANN) modelling techniques. The results showed that the models based on the hyperspectral images obtained in May achieved the best prediction, and the PLS-MLR model has a better stability and consistency than the PLS-ANN model. These results proviode an important theoretical and technical foundation for the future research and development of hyperspectral imaging-based citrus production techniques.